As a good Bayesian, I was considering a risk analysis as part of the complete consequences, but didn't make that clear, so thank you for clarifying that.
To carry this through using the argument of the study that finds a small correlation between vaccines and autism: if doctors lie about this, there's no guarantee the information about the study won't come out anyway, and then it will be known that they lied about them. That, more than the results themselves, would add real credibility to the conspiracy theory that doctors are covering up a connection between vaccines and autism. And worse, it would undermine the credibility of doctors with regards to other things, unrelated to vaccines.
The OP isn't considering the full consequences of undermining the credibility of the medical field to promote one political agenda, no matter how correct that political agenda is. In part, this is because the OP is assuming covering up this information would even work, and it very well might not.
At a more fundamental level, it would be important to release this information because science needs a mechanism for discovering when we're wrong or when things change. Right now, there's strong evidence for the null hypothesis: we can confidently say that vaccines do not cause autism. But if studies started finding a correlation between vaccines and autism, we'd want to know that. Maybe the diagnostic criteria for autism improves, maybe a new chemical is introduced to vaccines by a new manufacturing process, maybe a gene becomes more prevalent that interacts with existing vaccines; the possibilities are endless. Right now we should operate on the belief that vaccines don't cause autism, because that's what the information we have indicates, but if new information indicates something else, we should by all means update our beliefs and our actions.